Return-Path: 
 <users-return-3005-apmail-trafficserver-users-archive=trafficserver.apache.org@trafficserver.apache.org>
X-Original-To: apmail-trafficserver-users-archive@www.apache.org
Delivered-To: apmail-trafficserver-users-archive@www.apache.org
Received: from mail.apache.org (hermes.apache.org [140.211.11.3])
	by minotaur.apache.org (Postfix) with SMTP id CC9FE10CD7
	for <apmail-trafficserver-users-archive@www.apache.org>;
 Fri,  2 Aug 2013 11:08:26 +0000 (UTC)
Received: (qmail 86764 invoked by uid 500); 2 Aug 2013 11:08:25 -0000
Delivered-To: apmail-trafficserver-users-archive@trafficserver.apache.org
Received: (qmail 86745 invoked by uid 500); 2 Aug 2013 11:08:24 -0000
Mailing-List: contact users-help@trafficserver.apache.org; run by ezmlm
Precedence: bulk
List-Help: <mailto:users-help@trafficserver.apache.org>
List-Unsubscribe: <mailto:users-unsubscribe@trafficserver.apache.org>
List-Post: <mailto:users@trafficserver.apache.org>
List-Id: <users.trafficserver.apache.org>
Reply-To: users@trafficserver.apache.org
Delivered-To: mailing list users@trafficserver.apache.org
Received: (qmail 86729 invoked by uid 99); 2 Aug 2013 11:08:23 -0000
Received: from nike.apache.org (HELO nike.apache.org) (192.87.106.230)
    by apache.org (qpsmtpd/0.29) with ESMTP; Fri, 02 Aug 2013 11:08:23 +0000
X-ASF-Spam-Status: No, hits=1.5 required=5.0
	tests=HTML_MESSAGE,RCVD_IN_DNSWL_LOW,SPF_PASS
X-Spam-Check-By: apache.org
Received-SPF: pass (nike.apache.org: domain of yunkai.me@gmail.com designates
 209.85.215.173 as permitted sender)
Received: from [209.85.215.173] (HELO mail-ea0-f173.google.com)
 (209.85.215.173)
    by apache.org (qpsmtpd/0.29) with ESMTP; Fri, 02 Aug 2013 11:08:17 +0000
Received: by mail-ea0-f173.google.com with SMTP id g10so256835eak.32
        for <multiple recipients>; Fri, 02 Aug 2013 04:07:56 -0700 (PDT)
DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/relaxed;
        d=gmail.com; s=20120113;
        h=mime-version:in-reply-to:references:date:message-id:subject:from:to
         :cc:content-type;
        bh=0uufsZwj7LAsryhRxpLFtrogWbCIIbm75CSJXtJgAKM=;
        b=KyNUOufZCXgk4syqvaOJM70dSc/5fsVvZMSgp202NZZjgVVdUBTSDufUhFoTNV/tpp
         rhCuZVUMJX+pvIcUquIfRUYr3yvY12uaLGMJ9h9GwTJEq45IxT07J7Ijz40B1N/yBlw/
         E6vMCIxC8cK7reMJbTDkYNNvGPXLhYnUh7wsdd7cP3LHGtpR8qqO1x2FIIXl5biELddK
         ryb2BuqfexygD+n+LfWT+XUE3Dkx9pkJrLR8mCZySj14ee2n/8yiyXhAvm3WqlhGl4VY
         +uSJaFZLj4QqBkGTNWWHesZxcfJONgLc2NIdfozmdTU+cdUzwbUXxAwajIYGKPcdF0g+
         3WTA==
MIME-Version: 1.0
X-Received: by 10.15.90.139 with SMTP id q11mr5272249eez.137.1375441676826;
 Fri, 02 Aug 2013 04:07:56 -0700 (PDT)
Received: by 10.14.206.132 with HTTP; Fri, 2 Aug 2013 04:07:56 -0700 (PDT)
In-Reply-To: <C3E4DAA1-E8C3-49AC-A45C-EE0EDD3D8872@ogre.com>
References: <51EFF6CC.3080406@apache.org>
	<CABC3qWcz1h-2-xNjrgBNcFqahrrDT44ZbAQ+svx9CjonBSU2yw@mail.gmail.com>
	<26E7E8BA-D410-47BF-A334-037402375A1A@apache.org>
	<2606A17B-2598-4175-B8B7-DAD02CCD0F3E@gmail.com>
	<CAGEauyTdjmnhTrmnO4kRkHNhRfyq5h1sqt6m9hHdYEBbxS8AFQ@mail.gmail.com>
	<E5966E92-9B82-4357-9119-6D6EA8F2F1D1@apache.org>
	<A56F0026-D685-4E4C-834A-73FD3598662F@gmail.com>
	<C3E4DAA1-E8C3-49AC-A45C-EE0EDD3D8872@ogre.com>
Date: Fri, 2 Aug 2013 19:07:56 +0800
Message-ID: 
 <CAGEauyRY6UXYhCAdRgevq1CcfW98eQXd699e0W-WrjzqwrGyuQ@mail.gmail.com>
Subject: Re: [DISCUSS] Updates to records.config (default) for v3.4.0
From: Yunkai Zhang <yunkai.me@gmail.com>
To: dev-trafficserver <dev@trafficserver.apache.org>
Cc: "users@trafficserver.apache.org" <users@trafficserver.apache.org>
Content-Type: multipart/alternative; boundary=089e016351b68ea8c704e2f4f86f
X-Virus-Checked: Checked by ClamAV on apache.org

--089e016351b68ea8c704e2f4f86f
Content-Type: text/plain; charset=ISO-8859-1

On Fri, Aug 2, 2013 at 6:15 PM, Leif Hedstrom <leif@ogre.com> wrote:

>
> On Aug 2, 2013, at 12:01 PM, Yongming Zhao <ming.zym@gmail.com> wrote:
>
> > I'd like to enable that reclaimable free list by default building &
> installation, can we?
>
> I'd prefer not for v3.3.5 / 3.4.0. If you still do, I'd absolutely insist
> on it being off by default. We do not have enough data points for the
> usefulness of this feature, and what impact it might have on other systems
> / platforms / services (or e.g. NUMA).
>
> As you know, I still don't buy the reasons why reclaiming memory is
> necessary. It feels like a duct tape solution to an unknown problem (why is
> it allocating more memory than you want it to? Why are you running other
> applications / tools that are competing for memory? How do you avoid for
> the two to not compete for the same memory?).
>


The reason of why reclaming memory is need is that:

In the freelist, there are different size blocks:

1KB, 2KB, 4KB, 8KB, ...

different size block will be placed into different list.

suppose at some time, 2KB block is hot, so ATS will allocate many 2KB
blocks and put them into 2KB-list.

After a while, 4KB block becomes hot, then ATS will allocate many 4KB
blocks and put them into 4KB-list.

... 8KB ...
... 16KB ...
.... .... ....

After ATS running a long time, the number of allocated blocks in each list
will reach the *peek* level, but at some time, *not all* block is hot!

The blocks which not hot, but have been allocated at *peek* level, is the
root cause of memory wast.

That is why reclaiming memory is necessary for ATS, and why
reclaimable-freelist patch can take effect.



>
> All my $.01 of course.
>
> -- Leif
>
>


-- 
Yunkai Zhang
Work at Taobao

--089e016351b68ea8c704e2f4f86f
Content-Type: text/html; charset=ISO-8859-1
Content-Transfer-Encoding: quoted-printable

<div dir=3D"ltr"><br><div class=3D"gmail_extra"><br><br><div class=3D"gmail=
_quote">On Fri, Aug 2, 2013 at 6:15 PM, Leif Hedstrom <span dir=3D"ltr">&lt=
;<a href=3D"mailto:leif@ogre.com" target=3D"_blank">leif@ogre.com</a>&gt;</=
span> wrote:<br>
<blockquote class=3D"gmail_quote" style=3D"margin:0 0 0 .8ex;border-left:1p=
x #ccc solid;padding-left:1ex"><div class=3D"im"><br>
On Aug 2, 2013, at 12:01 PM, Yongming Zhao &lt;<a href=3D"mailto:ming.zym@g=
mail.com">ming.zym@gmail.com</a>&gt; wrote:<br>
<br>
&gt; I&#39;d like to enable that reclaimable free list by default building =
&amp; installation, can we?<br>
<br>
</div>I&#39;d prefer not for v3.3.5 / 3.4.0. If you still do, I&#39;d absol=
utely insist on it being off by default. We do not have enough data points =
for the usefulness of this feature, and what impact it might have on other =
systems / platforms / services (or e.g. NUMA).<br>

<br>
As you know, I still don&#39;t buy the reasons why reclaiming memory is nec=
essary. It feels like a duct tape solution to an unknown problem (why is it=
 allocating more memory than you want it to? Why are you running other appl=
ications / tools that are competing for memory? How do you avoid for the tw=
o to not compete for the same memory?).<br>
</blockquote><div><br></div><div><br></div><div>The reason of why reclaming=
 memory is need is that:</div><div><br></div><div>In the freelist, there ar=
e different size blocks:</div><div><br></div><div>1KB, 2KB, 4KB, 8KB, ...</=
div>
<div><br></div><div>different size block will be placed into different list=
.</div><div><br></div><div>suppose at some time, 2KB block is hot, so ATS w=
ill allocate many 2KB blocks and put them into 2KB-list.</div><div><br>
</div><div>After a while, 4KB block becomes hot, then ATS will allocate man=
y 4KB blocks and put them into 4KB-list.</div><div><br></div><div>... 8KB .=
..<br></div><div>... 16KB ...</div><div>.... .... ....</div><div><br></div>
<div>After ATS running a long time, the number of allocated blocks in each =
list will reach the *peek* level, but at some time, *not all* block is hot!=
=A0</div><div><br></div><div>The blocks which not hot, but have been alloca=
ted at *peek* level, is the root cause of memory wast.</div>
<div><br></div><div>That is why reclaiming memory is necessary for ATS, and=
 why reclaimable-freelist patch can take effect.=A0</div><div><br></div><di=
v>=A0</div><blockquote class=3D"gmail_quote" style=3D"margin:0 0 0 .8ex;bor=
der-left:1px #ccc solid;padding-left:1ex">

<br>
All my $.01 of course.<br>
<span class=3D"HOEnZb"><font color=3D"#888888"><br>
-- Leif<br>
<br>
</font></span></blockquote></div><br><br clear=3D"all"><div><br></div>-- <b=
r>Yunkai Zhang<div>Work at Taobao</div>
</div></div>

--089e016351b68ea8c704e2f4f86f--